{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","research_id":"eoa_inverse_innovation_exp09_light_prior_art_20260804","cell_id":"preimage_set_characterization__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["feature flag evaluate all combinations same output collision inverse mapping renderer","feature flag inverse mapping all configurations same variation output"],"source_ids":["SRC1","SRC2","SRC4"],"no_result_note":"No exact public implementation of a checkout-specific, output-conditioned feature-flag preimage ledger with the proposed completeness and use-guardrail contract was found."},"synonyms_and_historical_terms":{"queries":["feature toggle decision table truth table combinations targeting rules fallback output","feature model enumerate all configurations satisfying output constraint SAT solver"],"source_ids":["SRC2","SRC4"],"no_result_note":null},"products_practices_and_standards":{"queries":["LaunchDarkly SDK evaluation reasons fallback targeting rule fallthrough feature flags","NIST ACTS feature flag combinatorial testing constraints coverage"],"source_ids":["SRC1","SRC3"],"no_result_note":null},"component_combination":{"queries":["feature flag combinatorial interaction testing all configurations","projected model enumeration model counting satisfying assignments configuration space"],"source_ids":["SRC2","SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Flag evaluation rules in server-side SDKs","publisher":"LaunchDarkly","url":"https://launchdarkly.com/docs/sdk/concepts/flag-evaluation-rules","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["Feature-flag evaluation depends on context, prerequisites, individual targets, ordered targeting rules, fallthrough behavior, flag state, and SDK availability.","Distinct evaluation paths can return the same variation: offline and error conditions return a fallback; an off flag returns its off variation; prerequisite failure returns an off variation; and targeting or fallthrough can select configured variations.","LaunchDarkly can return an evaluation reason for one evaluation, providing a close rival to per-request tracing but not a complete inverse set over a declared domain."]},{"source_id":"SRC2","title":"Exploring Differences and Commonalities between Feature Flags and Configuration Options","publisher":"IEEE/ACM ICSE-SEIP","url":"https://cmustrudel.github.io/papers/meinicke20flags.pdf","source_type":"PRIMARY_RESEARCH","claims_supported":["Interview evidence indicates that feature-flag configuration knowledge and dependency documentation often rely on ad hoc mechanisms.","The paper reports that interacting flags and complicated data flows are difficult to reason about because of combinatorial growth in configurations and paths.","Existing configurable-system research includes feature models, constraint checking, interaction analysis, and configuration tracing, but interviewed feature-flag practitioners were generally unaware of comparable analysis tools for flags."]},{"source_id":"SRC3","title":"Combinatorial Methods for Trust and Assurance: Downloadable Tools","publisher":"National Institute of Standards and Technology","url":"https://csrc.nist.gov/Projects/Automated-Combinatorial-Testing-for-Software/Downloadable-Tools","source_type":"OFFICIAL_GUIDANCE","claims_supported":["NIST's ACTS generates constrained, variable-strength t-way test sets over input parameters.","NIST also provides coverage measurement and fault-identification tools for finding coverage gaps and likely fault-triggering combinations.","These practices are close to the proposal's finite-domain testing and collision analysis, but their stated contract is combinatorial coverage or fault identification rather than enumeration of every input producing one selected output."]},{"source_id":"SRC4","title":"Model Counting and Enumeration","publisher":"LogicNG","url":"https://logicng.org/documentation/model-counting-enumeration/","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["LogicNG supports enumerating every satisfying assignment of a formula and counting satisfying assignments.","It supports projected model enumeration over selected variables, closely matching the proposal's symbolic backsolve and multiplicity components.","The documented facility does not itself freeze and execute a feature-flag evaluator plus renderer, register operational boundary cases, or impose the proposal's advisory-use guardrails."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The technical ambiguity is visible: official evaluator documentation shows that flag state, context, prerequisites, ordered targeting rules, fallthrough, and availability/error paths can converge on the same returned variation, while primary research reports weak dependency documentation and combinatorial reasoning difficulty in feature-flag practice. The retained sources do not directly establish the proposal's checkout-dashboard data loss or that maintainers actually make unsafe rollback or deletion decisions from that output alone.","source_ids":["SRC1","SRC2"]},"closest_prior_art":[{"name":"LogicNG model enumeration and projected model enumeration","source_ids":["SRC4"],"overlap":"Enumerates all satisfying assignments, optionally projected onto selected variables, and supports model counting—the mathematical core of a finite output preimage and multiplicity ledger.","remaining_difference":"It is a general constraint-library capability, not a documented operational artifact that composes a frozen flag evaluator and renderer, records concrete boundary witnesses and indeterminate cases, labels the outside-domain region unknown, and restricts change authority."},{"name":"NIST ACTS and combinatorial coverage tooling","source_ids":["SRC3"],"overlap":"Models finite parameter domains and constraints, generates systematic combinations, measures coverage, and helps isolate influential combinations.","remaining_difference":"ACTS primarily supplies t-way covering samples and fault-oriented analysis; it does not promise every-and-only tuple yielding one specified rendered output or a complete target-fiber count."},{"name":"LaunchDarkly evaluation details and evaluation reasons","source_ids":["SRC1"],"overlap":"Exposes the evaluator's precedence paths and can distinguish why a single context received a variation, including off, prerequisite, target, rule, fallthrough, and error cases.","remaining_difference":"It explains or traces individual evaluations rather than characterizing all flag, context, and availability tuples compatible with a downstream rendered signature."},{"name":"Feature-model constraint and traceability practices","source_ids":["SRC2","SRC4"],"overlap":"Configuration research models constraints, enumerates valid configurations, detects interactions, and traces configuration decisions through implementations.","remaining_difference":"The retained evidence does not show an established feature-flag practice that conditions enumeration on one exact renderer output while also preserving witnesses, completeness accounting, boundary cases, unknown regions, and non-causality guardrails."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"For a frozen feature-flag evaluator and renderer over the declared 192-tuple domain, the proposed artifact can be distinguished by a single auditable contract: return every and only tuple producing the exact target variant and render signature, report target multiplicity and designated boundary cases, account for all 192 evaluations including indeterminate results, label all outside-domain states unknown, and forbid treating compatibility as actual causation or change authorization. The retained prior art supplies evaluator explanations, sampled interaction coverage, or generic satisfying-assignment enumeration, but no source demonstrates that complete operational combination.","contrastive_claim_falsifier":"The claim is falsified if an existing feature-flag or configuration-analysis facility is shown to consume the frozen evaluator-and-renderer semantics and already emit the complete exact-output preimage, multiplicity, boundary and indeterminate accounting, explicit outside-domain status, reproducible witnesses, and equivalent advisory-use restrictions; it is also empirically falsified if the pilot omits a ground-truth satisfying tuple, includes a nonsatisfying tuple, or declares completeness with any unevaluated tuple.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The bounded search covered direct phrasing, feature-toggle and configuration-model terminology, first-party evaluator behavior, official combinatorial-testing practice, and constraint-solver/model-enumeration components. Exactly four opened sources from four publishers were retained, including primary research, official guidance, and first-party documentation.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"The many-to-one technical mechanism and difficulty of reasoning across flag combinations are supported, although the checkout-specific dashboard and unsafe-decision frequency remain unverified; therefore problem evidence is partly rather than fully supported.","source_ids":["SRC1","SRC2"]},"distinct_testable_claim":{"status":"PASS","rationale":"The remaining claim is narrower than generic tracing, t-way testing, or model enumeration and has measurable set equality, multiplicity, boundary-accounting, and completeness conditions.","source_ids":["SRC1","SRC3","SRC4"]},"bounded_next_test":{"status":"PASS","rationale":"A hermetic 192-tuple run against frozen versions is finite and directly testable. Compare its returned set with an independently implemented exhaustive oracle, require 192 determinate outcomes, replay witnesses, and check seeded precedence, missing-attribute, and fallback boundaries before stopping.","source_ids":["SRC1","SRC3","SRC4"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"The proposed first test is read-only, synthetic, nonproduction, and reversible; it retains human deployment authority and includes appropriate halts for version mismatch, incomplete evaluation, sensitive data, or disagreement between symbolic and concrete evaluation. No retained source reveals an obvious prohibition on such sandbox evaluation.","source_ids":["SRC1","SRC3"]}},"screen_survival":true,"world_novelty_boundary":"This bounded public-web screen found adjacent evaluator-tracing, combinatorial-testing, feature-model, and satisfying-assignment-enumeration art, but no exact operational combination. It cannot establish world novelty, patentability, market size, expert acceptance, production prevalence, or realized safety or business value."}